Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add OneWave-AI/open-agent-stack --skill demo-videogit clone --depth 1 https://github.com/OneWave-AI/open-agent-stackWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/onewave-ai/open-agent-stack/demo-video)<a href="https://agentmods.dev/skills/onewave-ai/open-agent-stack/demo-video"><img src="https://agentmods.dev/badge/skills/onewave-ai/open-agent-stack/demo-video.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00067 | $0.01227 |
| Opus 5 | $0.00034 | $0.00613 |
| Sonnet 5 | $0.00013 | $0.00245 |
| Haiku 4.5 | $0.00007 | $0.00123 |
Grade A, and why
demo-video scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Demo Video
You shipped it and now something has to show it. The default result is a four-minute screen recording that opens with "so, first I'll log in" and loses everyone before the product does anything interesting.
The fix is a beat sheet before a single frame is captured. This skill plans and captures; rendering and editing belong to whatever video tooling the user already has.
The shape
Length by purpose, and none of these are negotiable upward:
| Purpose | Length | Opens with |
|---|---|---|
| Landing page hero loop | 10-20s, silent, looping | The product mid-action, no UI chrome |
| Launch post / social | 30-60s | The result, then how |
| Full demo / client walkthrough | 90-150s | The problem in one sentence |
Silent-by-default for anything embedded on a page: it autoplays muted, so it must work with no audio. Captions if there are words.
Workflow
-
Name the one thing. What single capability makes someone want this? The video shows that, well, and skips everything else. A demo covering six features communicates none of them. Everything that is not the one thing is a candidate for deletion, including login, settings, and navigation.
-
Write the beat sheet before capturing. Each beat gets one line of what is on screen and one line of what is said or shown. For a 60-second launch video that is roughly six beats:
- Hook (0-5s) -- the result, or the problem stated in a sentence. Never a logo, never "hi, I'm going to show you". The first three seconds decide whether there is a fourth.
- Context (5-12s) -- what this is and who it is for, once, plainly.
- The demo (12-45s) -- the one thing, actually working. Real data, real latency or an honest cut, real output.
- Proof (45-52s) -- the result on screen: the record created, the file produced, the time saved.
- Close (52-60s) -- what to do next and where. One action.
Read the beat sheet aloud against a timer before capturing. Overruns get cut here, where cutting is free.
-
Prepare the app for the camera. This is where most demo videos are lost:
- Seed realistic data. Empty states and
test test testmake a product look unused; obviously fake names make it look like a mockup. - Clean browser: no bookmarks bar, no extensions, no other tabs, no notifications. Use a fresh profile.
- Fixed window size at a 16:9 ratio, and set the zoom so text is readable at the size it will actually be watched -- usually 125-150%, because a 100% screenshot is illegible in a social feed.
- Hide anything private: real customer names, emails, keys, account balances, internal URLs. Check the browser autocomplete dropdown, which is a classic leak.
- Decide cursor treatment before starting. A visible click indicator helps; a wandering cursor does not.
- Seed realistic data. Empty states and
-
Capture in takes, not one pass. One beat per take. A flubbed beat costs one retake instead of the whole video. Move deliberately -- slightly slower than feels natural, with a beat of stillness before and after each action, which is what gives the edit somewhere to cut. Never capture typing in real time; type short, and speed it up later.
-
Handle the waiting honestly. Real latency is the thing videos most often fake. Cut the wait, or speed it up visibly, but do not present a four-second load as instant. Overstating speed is the demo lie that gets noticed on day one of the trial.
-
Narrate, or do not. If there is voiceover, write it as speech and record it separately from the capture -- narrating live produces both worse audio and worse pacing. If there is no voiceover, on-screen text carries the beats: short, high contrast, on screen long enough to read twice at a glance.
-
Verify where it will actually play. Watch it muted. Watch it on a phone. Watch the first three seconds alone and ask whether they earn the fourth. Check that the poster frame -- the still shown before play -- is a good frame, because for many viewers it is the only frame.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 6d ago First seen · 58 lines · 67 tokens per session scan A 21321823bfe4
demo-video is a skill published in the GitHub repository OneWave-AI/open-agent-stack (2 stars, last pushed 26d ago), licensed MIT. It adds 67 tokens to every session and 1,227 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
p5js
Use when users request: p5.js sketches, creative coding, generative art, interactive visualizations, canvas animations, browser-based visual art, data viz, shader effects, or any p5.js project.
audiocraft-audio-generation
AudioCraft: MusicGen text-to-music, AudioGen text-to-sound.
baoyu-comic
A creator for educational comics, including biographies and tutorials, that turns supplied content or topics into illustrated comic pages.
concept-diagrams
Generate flat, minimal educational SVG visuals as HTML.
hyperframes
Render MP4/WebM videos from HTML compositions.
touchdesigner-mcp
Control TouchDesigner via twozero MCP.